Reinforcement Learning For Risk-Averse Digital Asset Allocation

In recent years,application of reinforcement learning in Digital Asset Allo-cation and portfolio management has attracted more and more attention.However,most classic reinforcement learning algorithms ignore risk,which makes asset highly volatile and may lead to high max drawdown.In a real capital mar-ket,a investor will quit instead of holding on if the max drawdown is beyond his limit.According to this fact,we incorporates risk to the classical reinforce-ment learning framework by adjusting the reward with a risk penalty of max drawdown.